Meriem-DH/marine-dataset-cpt

---
license: cc-by-4.0
task_categories:
- text-generation
language:
- en
tags:
- ocean
- marine
- biology
pretty_name: Marine dataset
---
# Marine Biology - Continued Pre-Training Dataset

## Description
A corpus of Wikipedia articles covering marine biology and related domains, intended for continued pre-training (CPT) of language models on marine science knowledge.

## Content
Plain text articles scraped from Wikipedia across the following categories:
- Marine Biology
- Marine Ecology
- Ocean
- Coral Reefs
- Marine Mammals
- Oceanography
- Fisheries Science
- Marine Conservation

## Dataset Structure
| Split | Rows | Columns |
|-------|------|---------|
| train | 419 | title, text |
| test | 105 | title, text |

## Fields
- `title`: Wikipedia article title
- `text`: Clean plain text content of the article

## Construction
1. Article links scraped via Wikipedia Category API
2. Content fetched using Wikipedia API with `explaintext=True`
3. Text cleaned (whitespace normalization)
4. Split: 80% train / 20% test (seed=42)

## Intended Use
Continued pre-training phase before instruction fine-tuning.
Feed raw text to the model so it absorbs marine domain knowledge
before learning to answer questions.

## License
Wikipedia content is licensed under CC BY-SA 4.0.